Use hard and soft edges in a painting according to the job of each boundary: keep the sharpest useful contrast at the identity or focal anchors, soften transitions that turn a form, and lose boundaries where adjacent shapes can merge without confusing the subject. The goal is not “more blur.” It is a hierarchy in which a few edges ask for attention and the rest support them.
That hierarchy is easy to miss when painting from a photo. A camera may record a crisp hair silhouette, eyelash, nostril, collar, shoulder, and background seam at once. Copying all of them gives every boundary equal authority. Blurring all of them removes structure instead. The practical decision sits between those extremes.
What this four-pass test held still
We began with one fictional three-quarter portrait from an earlier FreePainter studio study. The sitter, crop, pose, expression, value structure, ochre blouse, neutral background, and upper-left light remain fixed. We made four digital edge passes:
- uniformly sharp: local contrast increased across the whole portrait;
- uniformly soft: the whole image received the same blur radius;
- camera copy: the source sharpness was left unchanged;
- designed hierarchy: the face stayed readable, the eye and mouth anchors remained found, and the hair, far cheek, shoulder, and background were selectively quieted.
These are controlled editorial simulations, not physical paintings and not a model benchmark. They test where edge contrast directs attention. They cannot predict how a particular brush, pigment, paper, drying time, or glaze will create a soft transition in real materials.
The cover makes the first result visible: all-sharp feels cut out, all-soft feels unfocused, and the unchanged source still gives incidental details too much weight. The designed pass is not “better” because it is smoother. It is clearer because edge strength follows importance.
The 1:3 label is a rehearsal constraint, not a universal measurement: reserve one edge family for the focal zone, then make the supporting families softer, lost, or merely found. It prevents “sharp” from becoming the default everywhere.
Four edge jobs—not four brush presets
The Getty’s formal-analysis guide notes that lines often define the edges of a form and lead the eye around a composition. That is the useful starting point: an edge is a relationship between neighboring shapes, not merely a brush setting.
Hard edge: identify or interrupt
A hard edge has an abrupt value or color change. In this portrait, the darkest eye accent against the lit face and the small mouth corner can stay firm because they help establish identity and expression. A cast shadow or an overlapping object may also need a clear break.
Failure mode: outlining the entire head, both lips, every curl, the neckline, and both shoulders with equal contrast. The portrait becomes a stack of stickers because no boundary is allowed to recede.
Soft edge: turn form without losing it
A soft edge changes gradually while the two neighboring regions remain distinguishable. The cheek turning toward shadow is the obvious case. A soft transition says that light is changing across one continuous volume rather than that two separate pieces meet.
The National Gallery of Art’s Renaissance teaching material describes Leonardo’s sfumato as gradual color change in place of precise contour, allowing facial forms to melt at their edges. That historical example establishes a real method, not a universal recipe. A portrait still needs some firmer anchors.
Failure mode: using the same soft brush everywhere. Uniform softness does not create form; it simply lowers information across the whole image.
Lost edge: merge on purpose
A lost edge occurs when neighboring shapes become close enough in value and color that part of their boundary disappears. Letting dark hair merge into a dark background can simplify the silhouette and make the lit face more active. Letting part of a shoulder dissolve can keep the lower frame from competing with the eyes.
The subject should remain reconstructable from the edges that stay found. If losing a boundary changes the head tilt, eye spacing, jaw width, or overlap of the neck and blouse, that edge was carrying structural evidence and should return.
Failure mode: treating “lost” as permission to erase a difficult drawing problem. Merge only after the geometry is correct.
Found edge: return only what the viewer needs
A found edge is the readable segment that resumes after a lost or softened passage. It may be a short hard accent or a modest value separation. In the cover, the far hair silhouette does not need to remain crisp around the whole head; one or two found sections are enough to confirm its shape.
Failure mode: restoring every softened boundary during the final detail pass. If all edges return, the hierarchy disappears with them.
The five-question edge audit
Run this audit at thumbnail size first, then at 100 percent:
- What is the focal zone? Name one compact area, not the entire subject. For a portrait, it may be the near eye, brow, nose bridge, and mouth relationship.
- Which three boundaries prove identity? Mark them before softening anything. Eye angle, jaw break, and hairline placement often matter more than individual lashes.
- Which transition describes one turning form? Soften that passage. The cheek, neck, and rounded shoulder should not read like separate cut-paper shapes.
- Which boundary can disappear without changing the subject? Test a hair-to-background or shoulder-to-background merge. Restore only enough to explain the silhouette.
- Where is the accidental sharpest edge? A bright collar seam or background stroke can steal focus. Lower its value contrast, soften it, or crop it before sharpening the face.
Do not count pixels or prescribe a fixed blur radius. Edge strength depends on output size, surrounding contrast, and viewing distance. Adobe’s imaging documentation is useful here because it separates two operations that are often confused: blur reduces sharpness by smoothing transitions, while sharpening creates the appearance of clarity by increasing contrast where tones meet. Neither operation decides which edge deserves attention. That remains a composition choice.
A reproducible two-minute test
Duplicate the same image four times. Do not change crop, exposure, color, or value. On the first copy, increase edge contrast everywhere. On the second, apply one even blur. Leave the third untouched. On the fourth, keep three identity anchors crisp, soften one form transition, and lose one background boundary.
Compare the four at the final delivery size. Ask where your eye lands first, whether the likeness survives, whether the face turns in space, and whether any peripheral boundary is louder than the focal zone. If the designed version works only when enlarged, the small-scale hierarchy is still too weak.
Then carry that edge map into one repaint. The FreePainter portrait-into-painting workspace is the deepest matching route: use one source and keep pose, crop, expression, and light fixed while you change the edge instruction. Judge likeness before texture. If the sitter changes, edge polish cannot rescue the experiment.
Limitations
Our SVG comparison uses deterministic digital blur and contrast to isolate edge hierarchy. Physical soft edges can come from wet-into-wet handling, scumbling, glazing, feathering, broken coverage, or adjacent mixtures with close values. Those methods leave different surfaces, and this test does not rank them.
The test also uses a portrait, where identity anchors make the cost of a misplaced edge obvious. Landscapes, geometric abstraction, ink drawing, and deliberately hard-edge painting may choose a different hierarchy. The transferable rule is narrower: make edge contrast intentional, and spend the sharpest differences where they explain the picture.
References
The sources below define the visual and imaging operations used in the audit. The four-pass comparison, the 1:3 working rule, and the decision map are original FreePainter Studio Research artifacts rather than measurements of historical paintings or physical media.